Image fusion is the process of combining multiple images of the same scene. These images can be acquired from different sensors, captured at different times, or may be having spectral and spatial characteristics. The resultant fused image will be having the most desirable characteristics of each image and will retain more relevant information in a single Image. It reduces redundancy and minimizes uncertainty and gets all the useful information 11.Image fusion must be robust, reliable, tolerant to noise or misregistrations. Generally fusion algorithms can be categorized into spatial domain fusion and transform domain fusion. The Spatial domain fusion directly deals with the pixels of the image. The pixel values are calculated by the use of the local spatial features such as gradient, variance, mean, standard deviation and spatial frequency to attain the desired result. The spatial domain fusion techniques are less complex, simple but not robust. In transform domain fusion, the images are first transferred in to frequency domain. Because of this, transform domain fusion techniques are used to represent the sharpness and edges of an image. The transformed coefficients provide the informations of an image which is used to select the blocks from source images to the fusion image34.
Figure 1 Generic method of transform based fusion
Categorization of image fusion techniques
Image fusion can be commonly categorised at four different stages: signal level, pixel level, feature level, and decision level.
1. Signal level fusion: Signals from different sensors are combined to get a new signal with a better Signal to Noise Ratio (SNR) than the original signals.
2. Pixel level fusion: It is performed on pixel by pixel basis to improve the quality of the fused image which gives information associated with each pixel.
3. Feature level fusion: It extracts the salient features such as pixel intensities, textures or edges. These similar features are fused from input images.
4. Decision level fusion: It combines the information at a higher level of abstraction and merges the results from multiple algorithms to get a final fused decision. 78
Applications of Image Fusion
Image fusion is used in many fields of remote sensing such as classification, change detection and object identification.
Classification of land cover features like water, Soil, forest and vegetation are the key tasks of remote sensing applications. When multiple accuracy of remote sensing images sources of images are fused, the classifications are enriched.
Change detection is an important process in monitoring and managing the natural resources. It is the process of identifying differences by observing it at different times. The merging of temporal images improves information on changes which is occurred in the observed area.18
Object Identification is used to identify the roads, buildings, bare soil ,grass land and tree using image fusion by maximize the amount of information extracted from satellite image data.
ANALYSIS OF VARIOUS FUSION METHODS
In general, Image fusion involves merging of two or more images in such a way to obtain the most desirable characteristics of each image. It is used to increase the reliability of the Interpretation. In order to detect changes in remote sensing multitemporal images, two different approaches are usually followed, one is with the fusion of images and the another one is without the fusion of images. Change detection without fusion results in loss of spatial and spectral information and leads to false positive results. On the other hand, change detection with fusion leads to provide more data for better interpretation. Considering the advantages of image fusion in change detection, the difference image obtained is fused to extract the inherent details for detecting changes.